DOCSLIB.ORG

Complexity and Algorithms for Nonlinear Optimization Problems

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Complexity and Algorithms for Nonlinear Optimization Problems Ann Oper Res (2007) 153: 257–296 DOI 10.1007/s10479-007-0172-6 Complexity and algorithms for nonlinear optimization problems Dorit S. Hochbaum Published online: 3 May 2007 © Springer Science+Business Media, LLC 2007 Abstract Nonlinear optimization algorithms are rarely discussed from a complexity point of view. Even the concept of solving nonlinear problems on digital computers is not well defined. The focus here is on a complexity approach for designing and analyzing algorithms for nonlinear optimization problems providing optimal solutions with prespecified accuracy in the solution space. We delineate the complexity status of convex problems over network constraints, dual of flow constraints, dual of multi-commodity, constraints defined by a sub- modular rank function (a generalized allocation problem), tree networks, diagonal dominant matrices, and nonlinear knapsack problem’s constraint. All these problems, except for the latter in integers, have polynomial time algorithms which may be viewed within a unifying framework of a proximity-scaling technique or a threshold technique. The complexity of many of these algorithms is furthermore best possible in that it matches lower bounds on the complexity of the respective problems. In general nonseparable optimization problems are shown to be considerably more dif- ficult than separable problems. We compare the complexity of continuous versus discrete nonlinear problems and list some major open problems in the area of nonlinear optimiza- tion. Keywords Nonlinear optimization · Convex network flow · Strongly polynomial algorithms · Lower bounds on complexity 1 Introduction Nonlinear optimization problems are considered to be harder than linear problems. This is the chief reason why approximate linear models are frequently used even if the circum- stances justify a nonlinear objective. A typical approach is to replace an objective function An earlier version of this paper appeared in 4OR, 3:3, 171–216, 2005. D.S. Hochbaum () Department of Industrial Engineering and Operations Research and Walter A. Haas School of Business, University of California, Berkeley, USA e-mail: [email protected] 258 Ann Oper Res (2007) 153: 257–296 that is nonlinear by a piecewise linear function. This approach may adversely affect the algo- rithm’s complexity as often the number of pieces is very large and integer variables may have to be introduced. For problems with nonlinear objective function the leading methodology consists of iterative and numerical algorithms in which complexity analysis is substituted with a convergence rate proof. In order to apply complexity analysis to nonlinear optimization problems, it is necessary to determine what it means to solve such a problem. Unlike linear problems, for nonlinear problems the length of the output can be infinite, such as in cases when a solution is irra- tional. There are two major complexity models for nonlinear optimization. One that seeks to approximate the objective function. The second, which we present here, approximates the optimal solution in the solution space. The latter has a number of advantages described next in Sect. 1.1. Our goals here are to set a framework for complexity analysis of nonlinear optimization with linear constraints, delineate as closely as possible the complexity borderlines between classes of nonlinear optimization problems, and generate a framework of effective tech- niques. In particular it is shown how properties of convexity, separability and quadraticness of nonlinear optimization contribute to a substantial reduction in the problems’ complexity. We review a spectrum of techniques that are most effective for each such class and demon- strate that in some cases the techniques lead to best possible algorithms in terms of their efficiency. This paper is an updated version of Hochbaum (2005). Some of the subjects covered appeared earlier in Hochbaum (1993). 1.1 The complexity model, or what constitutes a solution to a nonlinear optimization problem There is a fundamental difficulty in solving nonlinear optimization on digital computers. Un- like linear programming, for which all basic solutions to a set of linear inequalities require only finite accuracy of polynomial length in the length of the input (see, e.g., Papadimitiou and Steiglitz 1982, Lemma 2.1), one cannot bound a priori the number of digits required for the length of nonlinear programming optimal solutions. This feature is important since computers can only store numbers of finite accuracy. Even the simplest nonlinear optimiza- tion problems can have irrational solutions and thus writing the output alone requires infi- nite complexity. This is the case, for instance, in the minimization of the convex function max{x2 − 2, 0}. So the interpretation of what it means to solve a problem in reals is not obvious. Traditional techniques for coping with nonlinear problems are reviewed extensively in Minoux (1986). These techniques have several shortcomings. In some applications, the non- linear function is not available analytically. It is accessible via a data acquisition process or as a solution to a system of differential equations (as in some queuing systems). A typical traditional method approximates first the data input function as an analytic function, while assuming it is a polynomial of a conveniently low degree. Consequently, there are errors attributed to the inaccuracy of the assumed input even before an algorithm is applied for solving the problem. Further difficulties arise because of conditions required for the algorithms to work. The algorithms typically make use of information about the function’s derivatives from numeri- cal approximations, which incorporates a further element of error in the eventual outcome. Moreover, certain assumptions about the properties of the nonlinear functions, such as dif- ferentiability and continuity of the gradients, are often made without any evidence that the “true” functions indeed possess them. Ann Oper Res (2007) 153: 257–296 259 In addition to the computational difficulties, the output for nonlinear continuous opti- mization problems can consist of irrational or even transcendental (non algebraic) numbers. Since there is no finite representation of such numbers, they are usually truncated to fit within the prescribed accuracy of the hardware and software. Complexity analysis requires finite length input and output which is of polynomial length as a function of the length of the input. Therefore the usual presentation of nonlinear prob- lems renders complexity analysis inapplicable. For this reason complexity theory has not addressed nonlinear problems, with the exception of some quadratic problems. To resolve this issue, the nonlinear functions can be given in form of a table or oracle. An oracle ac- cepts arguments of limited number of digits, and outputs the value of the function truncated to the prescribed number of digits. Since nonlinear functions cannot be treated with the same absolute accuracy as linear functions, the notion of approximate solutions is particularly important. One definition of a solution to a nonlinear optimization problem is one that approximates the objective func- tion. Among the major proponents of this approach are Nemirovsky and Yudin (1983)who chose to approximate the objective value while requiring some prior knowledge about the properties of the objective function. We observe that any information about the behavior of the objective at the optimum can always be translated to a level of accuracy of the solution vector itself (and vice versa). A detailed discussion of this point is provided in Hochbaum and Shanthikumar (1990). We thus believe that the interest of solving the optimization prob- lem is in terms of the accuracy of the solution rather than the accuracy of the optimal objec- tive value. We thus use a second definition of a solution is one that approximates the solution in the solution space, within a prescribed accuracy of . According to the concept of -accuracy of Hochbaum and Shanthikumar (1990)asolutionissaidtobe-accurate if it is at most () at a distance of (in the L∞ norm) from an optimal solution. That is, a solution, x is ∗ () ∗ -accurate if there exists an optimal solution x such that ||x − x ||∞ ≤ . is then said to be the accuracy required in the solution space. In other words, the solution is identical to 1 the optimum in O(log ) decimal digits. Both definitions of a solution overlap for nonlinear problems on integers in which case an accuracy of <1 is sufficient. The computation model here assumes the unit cost model, i.e. any arithmetic operation or comparison is counted as a single operation, even if the operands are real numbers. We 1 use here however only operands with up to O(log ) significant digits. A similar complexity model using -accuracy was used by Shub and Smale (1996); Rene- gar (1987) and others. These works, however, address algebraic problems, as opposed to optimization problems addressed here. 1.2 Classes of nonlinear problems addressed and some difficult cases A fundamental concept linking between the solutions to continuous and integer problems is that of proximity. Classes of convex optimization problems for which there is a “good” proximity with relative closeness of the continuous and integer solutions are solvable in polynomial time that depends on that distance. These classes include problems with a con- straint matrix that has small subdeterminants. A prominent example of such optimization problems is the convex separable network flow problem.
Load more

Recommended publications
  • Nonlinear Optimization Using the Generalized Reduced Gradient Method Revue Française D’Automatique, Informatique, Recherche Opé- Rationnelle
    REVUE FRANÇAISE D’AUTOMATIQUE, INFORMATIQUE, RECHERCHE OPÉRATIONNELLE.RECHERCHE OPÉRATIONNELLE LEON S. LASDON RICHARD L. FOX MARGERY W. RATNER Nonlinear optimization using the generalized reduced gradient method Revue française d’automatique, informatique, recherche opé- rationnelle. Recherche opérationnelle, tome 8, no V3 (1974), p. 73-103 <http://www.numdam.org/item?id=RO_1974__8_3_73_0> © AFCET, 1974, tous droits réservés. L’accès aux archives de la revue « Revue française d’automatique, in- formatique, recherche opérationnelle. Recherche opérationnelle » implique l’accord avec les conditions générales d’utilisation (http://www.numdam.org/ conditions). Toute utilisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fi- chier doit contenir la présente mention de copyright. Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques http://www.numdam.org/ R.A.LR.O. (8* année, novembre 1974, V-3, p. 73 à 104) NONLINEAR OPTIMIZATION USING THE-GENERALIZED REDUCED GRADIENT METHOD (*) by Léon S. LÀSDON, Richard L. Fox and Margery W. RATNER Abstract. — This paper describes the principles and logic o f a System of computer programs for solving nonlinear optimization problems using a Generalized Reduced Gradient Algorithm, The work is based on earlier work of Âbadie (2). Since this paper was written, many changes have been made in the logic, and significant computational expérience has been obtained* We hope to report on this in a later paper. 1. INTRODUCTION Generalized Reduced Gradient methods are algorithms for solving non- linear programs of gênerai structure. This paper discusses the basic principles of GRG, and constructs a spécifie GRG algorithm. The logic of a computer program implementing this algorithm is presented by means of flow charts and discussion.
    [Show full text]
  • Combinatorial Structures in Nonlinear Programming
    Combinatorial Structures in Nonlinear Programming Stefan Scholtes¤ April 2002 Abstract Non-smoothness and non-convexity in optimization problems often arise because a combinatorial structure is imposed on smooth or convex data. The combinatorial aspect can be explicit, e.g. through the use of ”max”, ”min”, or ”if” statements in a model, or implicit as in the case of bilevel optimization where the combinatorial structure arises from the possible choices of active constraints in the lower level problem. In analyzing such problems, it is desirable to decouple the combinatorial from the nonlinear aspect and deal with them separately. This paper suggests a problem formulation which explicitly decouples the two aspects. We show that such combinatorial nonlinear programs, despite their inherent non-convexity, allow for a convex first order local optimality condition which is generic and tight. The stationarity condition can be phrased in terms of Lagrange multipliers which allows an extension of the popular sequential quadratic programming (SQP) approach to solve these problems. We show that the favorable local convergence properties of SQP are retained in this setting. The computational effectiveness of the method depends on our ability to solve the subproblems efficiently which, in turn, depends on the representation of the governing combinatorial structure. We illustrate the potential of the approach by applying it to optimization problems with max-min constraints which arise, for example, in robust optimization. 1 Introduction Nonlinear programming is nowadays regarded as a mature field. A combination of important algorithmic developments and increased computing power over the past decades have advanced the field to a stage where the majority of practical prob- lems can be solved efficiently by commercial software.
    [Show full text]
  • Nonlinear Integer Programming ∗
    Nonlinear Integer Programming ∗ Raymond Hemmecke, Matthias Koppe,¨ Jon Lee and Robert Weismantel Abstract. Research efforts of the past fifty years have led to a development of linear integer programming as a mature discipline of mathematical optimization. Such a level of maturity has not been reached when one considers nonlinear systems subject to integrality requirements for the variables. This chapter is dedicated to this topic. The primary goal is a study of a simple version of general nonlinear integer problems, where all constraints are still linear. Our focus is on the computational complexity of the problem, which varies significantly with the type of nonlinear objective function in combination with the underlying combinatorial structure. Nu- merous boundary cases of complexity emerge, which sometimes surprisingly lead even to polynomial time algorithms. We also cover recent successful approaches for more general classes of problems. Though no positive theoretical efficiency results are available, nor are they likely to ever be available, these seem to be the currently most successful and interesting approaches for solving practical problems. It is our belief that the study of algorithms motivated by theoretical considera- tions and those motivated by our desire to solve practical instances should and do inform one another. So it is with this viewpoint that we present the subject, and it is in this direction that we hope to spark further research. Raymond Hemmecke Otto-von-Guericke-Universitat¨ Magdeburg, FMA/IMO, Universitatsplatz¨ 2, 39106 Magdeburg, Germany, e-mail: [email protected] Matthias Koppe¨ University of California, Davis, Dept. of Mathematics, One Shields Avenue, Davis, CA, 95616, USA, e-mail: [email protected] Jon Lee IBM T.J.
    [Show full text]
  • An Improved Sequential Quadratic Programming Method Based on Positive Constraint Sets, Chemical Engineering Transactions, 81, 157-162 DOI:10.3303/CET2081027
    157 A publication of CHEMICAL ENGINEERING TRANSACTIONS VOL. 81, 2020 The Italian Association of Chemical Engineering Online at www.cetjournal.it Guest Editors: Petar S. Varbanov, Qiuwang Wang, Min Zeng, Panos Seferlis, Ting Ma, Jiří J. Klemeš Copyright © 2020, AIDIC Servizi S.r.l. DOI: 10.3303/CET2081027 ISBN 978-88-95608-79-2; ISSN 2283-9216 An Improved Sequential Quadratic Programming Method Based on Positive Constraint Sets Li Xiaa,*, Jianyang Linga, Rongshan Bia, Wenying Zhaob, Xiaorong Caob, Shuguang Xianga a State Key Laboratory Base for Eco-Chemical Engineering, College of Chemical Engineering, Qingdao University of Science and Technology, Zhengzhou Road No. 53, Qingdao 266042, China b Chemistry and Chemistry Engineering Faculty,Qilu Normal University, jinan, 250013, Shandong, China [email protected] SQP (sequential quadratic programming) is an effective method to solve nonlinear constraint problems, widely used in chemical process simulation optimization. At present, most optimization methods in general chemical process simulation software have the problems of slow calculation and poor convergence. To solve these problems , an improved SQP optimization method based on positive constraint sets was developed. The new optimization method was used in the general chemical process simulation software named optimization engineers , adopting the traditional Armijo type step rule, L-BFGS (Limited-Memory BFGS) algorithm and rules of positive definite matrix. Compared to the traditional SQP method, the SQP optimization method based on positive constraint sets simplifies the constraint number of corresponding subproblems. L-BFGS algorithm and rules of positive definite matrix simplify the solution of the second derivative matrix, and reduce the amount of storage in the calculation process.
    [Show full text]
  • Appendix a Solving Systems of Nonlinear Equations
    Appendix A Solving Systems of Nonlinear Equations Chapter 4 of this book describes and analyzes the power flow problem. In its ac version, this problem is a system of nonlinear equations. This appendix describes the most common method for solving a system of nonlinear equations, namely, the Newton-Raphson method. This is an iterative method that uses initial values for the unknowns and, then, at each iteration, updates these values until no change occurs in two consecutive iterations. For the sake of clarity, we first describe the working of this method for the case of just one nonlinear equation with one unknown. Then, the general case of n nonlinear equations and n unknowns is considered. We also explain how to directly solve systems of nonlinear equations using appropriate software. A.1 Newton-Raphson Algorithm The Newton-Raphson algorithm is described in this section. A.1.1 One Unknown Consider a nonlinear function f .x/ W R ! R. We aim at finding a value of x so that: f .x/ D 0: (A.1) .0/ To do so, we first consider a given value of x, e.g., x . In general, we have that f x.0/ ¤ 0. Thus, it is necessary to find x.0/ so that f x.0/ C x.0/ D 0. © Springer International Publishing AG 2018 271 A.J. Conejo, L. Baringo, Power System Operations, Power Electronics and Power Systems, https://doi.org/10.1007/978-3-319-69407-8 272 A Solving Systems of Nonlinear Equations Using Taylor series, we can express f x.0/ C x.0/ as: Â Ã.0/ 2 Â Ã.0/ df .x/ x.0/ d2f .x/ f x.0/ C x.0/ D f x.0/ C x.0/ C C ::: dx 2 dx2 (A.2) .0/ Considering only the first two terms in Eq.
    [Show full text]
  • A Sequential Quadratic Programming Algorithm with an Additional Equality Constrained Phase
    A Sequential Quadratic Programming Algorithm with an Additional Equality Constrained Phase Jos´eLuis Morales∗ Jorge Nocedal † Yuchen Wu† December 28, 2008 Abstract A sequential quadratic programming (SQP) method is presented that aims to over- come some of the drawbacks of contemporary SQP methods. It avoids the difficulties associated with indefinite quadratic programming subproblems by defining this sub- problem to be always convex. The novel feature of the approach is the addition of an equality constrained phase that promotes fast convergence and improves performance in the presence of ill conditioning. This equality constrained phase uses exact second order information and can be implemented using either a direct solve or an iterative method. The paper studies the global and local convergence properties of the new algorithm and presents a set of numerical experiments to illustrate its practical performance. 1 Introduction Sequential quadratic programming (SQP) methods are very effective techniques for solv- ing small, medium-size and certain classes of large-scale nonlinear programming problems. They are often preferable to interior-point methods when a sequence of related problems must be solved (as in branch and bound methods) and more generally, when a good estimate of the solution is available. Some SQP methods employ convex quadratic programming sub- problems for the step computation (typically using quasi-Newton Hessian approximations) while other variants define the Hessian of the SQP model using second derivative informa- tion, which can lead to nonconvex quadratic subproblems; see [27, 1] for surveys on SQP methods. ∗Departamento de Matem´aticas, Instituto Tecnol´ogico Aut´onomode M´exico, M´exico. This author was supported by Asociaci´onMexicana de Cultura AC and CONACyT-NSF grant J110.388/2006.
    [Show full text]
  • GRASP for NONLINEAR OPTIMIZATION in This Paper, We
    GRASP FOR NONLINEAR OPTIMIZATION C.N. MENESES, P.M. PARDALOS, AND M.G.C. RESENDE ABSTRACT. We propose a Greedy Randomized Adaptive Search Procedure (GRASP) for solving continuous global optimization problems subject to box constraints. The method was tested on benchmark functions and the computational results show that our approach was able to find in a few seconds optimal solutions for all tested functions despite not using any gradient information about the function being tested. Most metaheuristcs found in the literature have not been capable of finding optimal solutions to the same collection of functions. 1. INTRODUCTION In this paper, we consider the global optimization problem: Find a point x∗ such that n f (x∗) f (x) for all x X, where X is a convex set defined by box constraints in R . For ≤ 2 instance, minimize f (x ;x ) = 100(x x2)2 +(1 x )2 such that 2 x 2 for i = 1;2. 1 2 2 − 1 − 1 − ≤ i ≤ Many optimization methods have been proposed to tackle continuous global optimiza- tion problems with box constraints. Some of these methods use information about the function being examined, e.g. the Hessian matrix or gradient vector. For some functions, however, this type of information can be difficult to obtain. In this paper, we pursue a heuristic approach to global optimization. Even though several heuristics have been designed and implemented for the problem discussed in this paper (see e.g. [1, 15, 16, 17]), we feel that there is room for yet an- other one because these heuristics were not capable of finding optimal solutions to several functions in standard benchmarks for continuous global optimization problems with box constraints.
    [Show full text]
  • Parallel Technique for the Metaheuristic Algorithms Using Devoted Local Search and Manipulating the Solutions Space
    applied sciences Article Parallel Technique for the Metaheuristic Algorithms Using Devoted Local Search and Manipulating the Solutions Space Dawid Połap 1,* ID , Karolina K˛esik 1, Marcin Wo´zniak 1 ID and Robertas Damaševiˇcius 2 ID 1 Institute of Mathematics, Silesian University of Technology, Kaszubska 23, 44-100 Gliwice, Poland; [email protected] (K.K.); [email protected] (M.W.) 2 Department of Software Engineering, Kaunas University of Technology, Studentu 50, LT-51368, Kaunas, Lithuania; [email protected] * Correspondence: [email protected] Received: 16 December 2017; Accepted: 13 February 2018 ; Published: 16 February 2018 Abstract: The increasing exploration of alternative methods for solving optimization problems causes that parallelization and modification of the existing algorithms are necessary. Obtaining the right solution using the meta-heuristic algorithm may require long operating time or a large number of iterations or individuals in a population. The higher the number, the longer the operation time. In order to minimize not only the time, but also the value of the parameters we suggest three proposition to increase the efficiency of classical methods. The first one is to use the method of searching through the neighborhood in order to minimize the solution space exploration. Moreover, task distribution between threads and CPU cores can affect the speed of the algorithm and therefore make it work more efficiently. The second proposition involves manipulating the solutions space to minimize the number of calculations. In addition, the third proposition is the combination of the previous two. All propositions has been described, tested and analyzed due to the use of various test functions.
    [Show full text]
  • A New Successive Linearization Approach for Solving Nonlinear Programming Problems
    Available at http://pvamu.edu/aam Applications and Applied Appl. Appl. Math. Mathematics: ISSN: 1932-9466 An International Journal (AAM) Vol. 14, Issue 1 (June 2019), pp. 437 – 451 A New Successive Linearization Approach for Solving Nonlinear Programming Problems 1Inci Albayrak, 2Mustafa Sivri and 3Gizem Temelcan 1,2 Department of Mathematical Engineering Yildiz Technical University Davutpasa, Istanbul, Turkey [email protected]; [email protected] 3 Department of Computer Programming Istanbul Aydin University Kucukcekmece, Istanbul, Turkey [email protected] Abstract In this paper, we focused on general nonlinear programming (NLP) problems having m nonlinear (or linear) algebraic inequality (or equality or mixed) constraints with a nonlinear (or linear) algebraic objective function in n variables. We proposed a new two-phase-successive linearization approach for solving NLP problems. Aim of this proposed approach is to find a solution of the NLP problem, based on optimal solution of linear programming (LP) problems, satisfying the nonlinear constraints oversensitively. This approach leads to novel methods. Numerical examples are given to illustrate the approach. Keywords: Nonlinear programming problems; Taylor series; Linear programming problems; Hessian matrix; Maclaurin series; Linearization approach MSC 2010 No.: 90C05, 41A58, 93B18 1. Introduction Optimization occurs in many fields. Constructing a mathematical model for real life problems is important for optimizers to find optimal strategies effectively. Optimization problems can be classified according to the nature of the objective function and constraints. An optimization problem can be defined as min (or max) of a single (or multi) objective function, subject to (or not to) single (or multi) nonlinear (or linear) inequality (or equality or mixed) constraints.
    [Show full text]
  • Constrained Optimization 5
    Constrained Optimization 5 Most problems in structural optimization must be formulated as constrained min- imization problems. In a typical structural design problem the objective function is a fairly simple function of the design variables (e.g., weight), but the design has to satisfy a host of stress, displacement, buckling, and frequency constraints. These constraints are usually complex functions of the design variables available only from an analysis of a finite element model of the structure. This chapter offers a review of methods that are commonly used to solve such constrained problems. The methods described in this chapter are for use when the computational cost of evaluating the objective function and constraints is small or moderate. In these meth- ods the objective function or constraints these are calculated exactly (e.g., by a finite element program) whenever they are required by the optimization algorithm. This approach can require hundreds of evaluations of objective function and constraints, and is not practical for problems where a single evaluation is computationally ex- pensive. For these more expensive problems we go through an intermediate stage of constructing approximations for the objective function and constraints, or at least for the more expensive functions. The optimization is then performed on the approx- imate problem. This approximation process is described in the next chapter. The basic problem that we consider in this chapter is the minimization of a function subject to equality and inequality constraints minimize f(x) such that hi(x) = 0, i = 1, . , ne , (5.1) gj(x) ≥ 0, j = 1, . , ng . The constraints divide the design space into two domains, the feasible domain where the constraints are satisfied, and the infeasible domain where at least one of the constraints is violated.
    [Show full text]
  • Tabu Search for Constraint Satisfaction
    Tabu Search for Constraint Solving and Its Applications Jin-Kao Hao LERIA University of Angers 2 Boulevard Lavoisier 49045 Angers Cedex 01 - France 1. Introduction The Constraint Satisfaction Problem (CSP) is a very general problem able to conveniently formulate a wide range of important applications in combinatorial optimization. These include well-known classical problems such as graph k-coloring and satisfiability, as well as many practical applications related to resource assignments, planning and timetabling. The feasibility problem of Integer Programming can also be easily represented by the CSP formalism. As elaborated below, we provide case studies for specific CSP applications, including the important realm of frequency assignment, and establish that tabu search is the best performing approach for these applications. The Constraint Satisfaction Problem is defined by a triplet (X, D, C) where: – X = {x1,· · ·,xn} is a finite set of n variables. – D = {Dx1 ,· · ·,Dxn} is a set of associated domains. Each domain Dxi specifies the finite set of possible values of the variable xi. – C = {C1,· · ·,Cp} is a finite set of p constraints. Each constraint is defined on a set of variables and specifies which combinations of values are compatible for these variables. Given such a triplet, the basic problem consists in finding a complete assignment of the values to the variables such that that all the constraints are satisfied. Such an assignment is called a solution to the given CSP instance. Since the set of all assignments is given by the Cartesian product Dx1 ×· · ·×Dxn, solving a CSP means to determine a particular assignment among a potentially huge search space.
    [Show full text]
  • Constrained Quadratic Optimization: Theory and Application for Wireless Communication Systems
    Constrained Quadratic Optimization: Theory and Application for Wireless Communication Systems by Erik Hons A thesis presented to the University of Waterloo in ful¯lment of the thesis requirement for the degree of Master of Applied Science in Electrical and Computer Engineering Waterloo, Ontario, Canada, 2001 c Erik Hons 2001 ° I hereby declare that I am the sole author of this thesis. I authorize the University of Waterloo to lend this thesis to other institutions or individuals for the purpose of scholarly research. I further authorize the University of Waterloo to reproduce this thesis by pho- tocopying or by other means, in total or in part, at the request of other institutions or individuals for the purpose of scholarly research. ii The University of Waterloo requires the signatures of all persons using or pho- tocopying this thesis. Please sign below, and give address and date. iii Abstract The suitability of general constrained quadratic optimization as a modeling and solution tool for wireless communication is discussed. It is found to be appropriate due to the many quadratic entities present in wireless communication models. It is posited that decisions which a®ect the minimum mean squared error at the receiver may achieve improved performance if those decisions are energy constrained. That theory is tested by applying a speci¯c type of constrained quadratic optimization problem to the forward link of a cellular DS-CDMA system. Through simulation it is shown that when channel coding is used, energy constrained methods typi- cally outperform unconstrained methods. Furthermore, a new energy-constrained method is presented which signi¯cantly outperforms a similar published method in several di®erent environments.
    [Show full text]